DraftKings’ AI Model Targets Loss‑Making Gamblers with Promotions
What Happened — Investigations by The New York Times and ProPublica reveal that DraftKings’ machine‑learning model scores customers on how likely they are to increase betting after receiving a promotion. The model uses betting frequency, loss amounts, and account balances to push targeted offers to users who are already losing money.
Why It Matters for Trust & Control Assurance
- This scenario tests the control objective of responsible AI governance – ensuring that automated decision‑making systems are designed, deployed, and monitored to avoid harmful outcomes.
- Continuous control‑assurance programs need documented policies, impact assessments, and monitoring evidence to demonstrate that AI models do not exploit vulnerable users.
- Mapping this to Verisq’s Control Mapping capability helps organizations prove alignment with AI‑risk frameworks (e.g., NIST AI RMF) across multiple compliance regimes.
Who Is Affected – Online gambling platforms, digital‑media advertisers, and any organization that uses AI‑driven personalization for high‑risk consumer segments.
Recommended Actions
- Conduct an AI‑risk impact assessment focused on consumer‑harm outcomes.
- Document model design decisions, data sources, and mitigation controls in a continuous‑evidence repository.
- Implement a monitoring process that flags “elasticity” scores exceeding predefined risk thresholds and triggers responsible‑use reviews.
Source: Malwarebytes Labs
Technical Notes
- The model ingests transactional betting data (frequency, loss amount, balance) to compute an “elasticity” score.
- No software vulnerability is disclosed; the risk stems from the model’s objective and deployment without adequate ethical safeguards.
Source: Malwarebytes Labs